The Function of Guanylate Cyclase 1 and Guanylate Cyclase 2 in Rod and Cone Photoreceptors*

Retinal guanylate cyclases 1 and 2 (GC1 and GC2) are responsible for synthesis of cyclic GMP in rods and cones, but their individual contributions to phototransduction are unknown. We report here that the deletion of both GC1 and GC2 rendered rod and cone photoreceptors nonfunctional and unstable. In the rod outer segments of GC double knock-out mice, guanylate cyclase-activating proteins 1 and 2, and cyclic GMP phosphodiesterase were undetectable, although rhodopsin and transducin α-subunit were mostly unaffected. Outer segment membranes of GC1–/– and GC double knock-out cones were destabilized and devoid of cone transducin (α- and γ-subunits), cone phosphodiesterase, and G protein-coupled receptor kinase 1, whereas cone pigments were present at reduced levels. Real time reverse transcription-PCR analyses demonstrated normal RNA transcript levels for the down-regulated proteins, indicating that down-regulation is posttranslational. We interpret these results to demonstrate an intrinsic requirement of GCs for stability and/or transport of a set of membrane-associated phototransduction proteins.

brane proteins with a single transmembrane domain (1)(2)(3). GC1 is detected in the retina, pineal gland, and olfactory bulb (4) as well as the cochlear nerve and the organ of Corti (5), whereas GC2 is found only in the retina. The activities of these enzymes are Ca 2ϩ -sensitive, a sensitivity that is mediated by guanylate cyclase-activating proteins (GCAPs). GCAPs are Ca 2ϩ -binding proteins with three high affinity Ca 2ϩ -binding sites (EF hands) (6 -8). In light, the activation of cGMP phosphodiesterase (PDE6) in rods and cones leads to the hydrolysis of cGMP and the closure of cGMP-gated (CNG) cation channels (for a review, see Ref. 9). Channel closure causes the cytosolic free Ca 2ϩ concentration to decline. The resulting dissociation of Ca 2ϩ from the GCAPs increases the activity of the GCs and therefore the cGMP level, completing the negative feedback loop (10). Thus, GCs are not only important for maintaining the basal cGMP level in darkness but are also critical for expediting the recovery of photoreceptors following light stimulation. Of the two GCAPs present in mouse retina, GCAP1 stimulates GC1 more efficiently than GC2 (11). Deletion of both GCAP genes delays recovery of the dark current due to loss of Ca 2ϩ -dependent GC activation (12).
In addition to particulate GCs, a role in photoreceptor physiology was envisioned for soluble GCs, but no genetic or functional evidence is available for a role in modulation of cGMP in photoreceptors (13,14). In humans, null mutations in the GC1 gene (GUCY2D, on chromosome 17) are known to cause Leber congenital amaurosis type 1 (LCA1), an autosomal recessive, early onset rod/cone dystrophy (15,16). A naturally occurring retinal degeneration in chicken (rd or GUCY1*B chicken), caused also by a null mutation in the GC1 gene (17), mimics the human LCA phenotype. In mice, however, a null mutation in the GC1 gene produces a cone dystrophy; the GC1 Ϫ/Ϫ rods remain viable and responsive to light (18,19). No human retinal disease has yet been linked to a defect in the GC2 gene (GUCY2F, on the X chromosome). Here, the functional role of GC2 was examined by deleting the mouse GC2 gene. Our results show that electroretinographic (ERG) responses are largely normal in GC2 Ϫ/Ϫ retina. In contrast, the ERG responses of GC1/GC2 double knock-out (GCdko) mice are absent, suggesting that no GC other than GC1 and GC2 is involved in rod or cone phototransduction. Histology of GCdko retinas revealed that outer segments form but degenerate in a manner similar to that observed in the rd (GUCY1*B) chicken (17) and in human LCA1 patients (20). Ultrastructural analyses detailed abnormal outer segment morphologies. Down-regula-tion of PDE6 and GCAPs in GCdko rods as well as down-regulation of cone phototransduction components in GC1 Ϫ/Ϫ cones suggest that GCs serve enzymatic, stabilizing, and structural/transport roles in photoreceptors.

Construction of Targeting Vector and Generation of GC2
Knock-out Mice-To delete the GC2 gene, a 16-kb mouse genomic DNA fragment containing exon 1 was cloned from a mouse 129Sv/Ev genomic library. An 8.5-kb XbaI fragment was subcloned to produce the targeting vector. One side of the neo gene cassette was inserted 5 bp upstream of the ATG start codon in exon 2. The replaced sequence starts with 5Ј-AGG-CTATGTTCCTCGGACCCTGGCCTTTTTCTCGCC. The other side of the neo gene cassette was inserted 290 bp downstream of the ATG start codon inside exon 2, and that replaced sequence reads 5Ј-ATTCGAAGGAGTAACTCCTGTCAAA-TGTCTTGTCCCGG. In this strategy, the neo cassette replaced part of the coding region of exon 2. Ten micrograms of the targeting vector were linearized by NotI and then transfected by electroporation of iTL1 ES cells (129Sv/Ev) embryonic stem cells. After selection in G418, surviving colonies were expanded; PCR analysis was performed using primer pair Gctoda6 and PLA2 to identify recombinant clones. Primer Gctoda6 is located 165 bp downstream of the XbaI site, on the short arm side, with a sequence of 5Ј-GTTCTGAGCTACAG-ATCCTACAGTG. Primer PLA2 is located in the 5Ј-promoter region of the neo gene cassette and has a sequence of 5Ј-GTT-CTTCGGACGCCTCGTCAACAC. The positive clones gave rise to a 1.5-kb PCR fragment. Correctly targeted ES cell lines were microinjected into C57BL/6J blastocysts. Chimeric mice were generated and gave germ line transmission of the GC2 knock-out to progeny. Primer pair Gctoda6 and Gctowt1 was used to identify the WT allele by its 1.3-kb amplified product. Primer Gctowt1 is located 280 bp downstream of the ATG start codon inside exon 1, inside the neo cassette replaced area, with a sequence of 5Ј-CAAGACATTTGACAGGAGTTACTC. Alternatively, primer Gctowt1 can be exchanged for primer Gctowt2, which is located 6 bp upstream of primer Gctowt1, with a sequence of 5Ј-CTGCTTTAGCAATTGAGCGAAT-CAG. Amplification yields a 1.3-kb fragment. PCR conditions were 94°C for 20 s, 62°C for 60 s, and 72°C for 120 s (35 cycles). In difficult amplifications, 10% Me 2 SO or Q solution from Qiagen was used.
GC1 Ϫ/Ϫ and Double Knock-out Mice-Procedures for the animal experiments described here were IACUC-approved by the University of Utah and Case Western Reserve University and conformed to recommendations of the Association of Research for Vision and Ophthalmology. Animals were maintained in complete darkness, or cyclic light (12 h light/12 h dark or 14 h light/10 h dark) conditions, and physiological experiments were performed under dim red illumination using a Kodak number 1 Safelight filter (transmittance Ͼ560 nm). GC1 knock-out mice were rederived from GC1 knock-outs (19) using mice originally produced by Dr. David Garbers (University of Texas Southwestern, Dallas, TX) (18). GC single and double knock-outs were typed for the presence of GC1 and GC2 knock-out alleles and absence of WT alleles. Primers used for GC1 genotyping were as follows: Gc1F4 (forward primer in intron 4), 5Ј-TCCTATCCACGACAGGACCAAGACTGT; Gc1R4 (reverse primer in intron 5), 5Ј-GAGAGCAGAAGGG-TAGCATTAGCTCAG; NeoF4 (forward primer in neo cassette), 5Ј-ACCGCTATCAGGACATAGCGTTGGCTA. Both pairs of primers were used in a one-tube PCR amplification to yield the fragments shown in Fig. 1D. GC2 Ϫ/Ϫ mice were genotyped using forward primer Pla2 and reverse primer Gc2mt-R2 (5Ј-GTGCCTACAGGACTCTTGATGTCATTC). WT mice were genotyped with primers Gctowt1 and Gctoda6.
Immunoblotting-Retinas from both eyes of the mouse were sonicated in 100 l of lysis buffer (26). Each sample (15 g of protein) was separated on a 10 -12% SDS-PAGE, transferred onto a polyvinylidene difluoride filter, and probed with primary antibodies followed by horseradish peroxidase-conjugated secondary antibody. The signal was visualized using chemiluminescence (ECL Plus system (Amersham Biosciences).
ERGs-The recording procedure was performed as described (27,28). ERGs were recorded with the universal testing and electrophysiologic system UTAS E-3000 (LKC Technologies, Inc.). The light intensity was calibrated by the manufacturer and computer-controlled. Flash stimuli had a range of intensities (Ϫ3.7-2.8 log cd⅐s⅐m Ϫ2 ), and white light flash duration was adjusted according to intensity (from 20 s to 1 ms). Leading edges of the ERG responses were fitted with a model of rod photoreceptor activation (29). The double-flash recording followed a previously published protocol (30). The normalized amplitude of the probe flash a-wave versus the time between two flashes was plotted and fit by the linear regression algorithm in the SigmaPlot program (version 9.0). Dark-adapted ERG recordings after intense constant illumination were performed and evaluated as described (28,31).
Single Cell Recordings-Single cell recordings were performed as described previously (32). An individual rod outer segment was drawn into a snug-fitting glass suction electrode containing 140 mM NaCl, 3.6 mM KCl, 2.4 mM MgCl 2 , 1.2 mM CaCl 2 , 3 mM HEPES, pH 7.4, 0.02 mM EDTA, and 10 mM glucose. Membrane current was monitored with a current-to-voltage amplifier (Axopatch 200B, Axon Instruments, CA). All signals were low pass-filtered at 20 Hz (8-pole Bessel) and sampled at 500 Hz. The light stimulus consisted of brief flashes (10 ms) of unpolarized 500-nm (10-nm bandwidth) light with intervals of 8 s.
Histology-Right eyes of 1-2-month-old WT, GC1 Ϫ/Ϫ , GC2 Ϫ/Ϫ , and GCdko mice were prepared for light microscope plastic sections and electron microscope thin sections as described (33). For electron microscopy, the eyecups were postfixed in 1% osmium tetroxide in phosphate buffer for 1 h, dehydrated through an ascending series of graded ethanols, and embedded in Spurr's resin. Sections 0.5-1 m thick and passing through the optic nerve were imaged using a Leica DM-R microscope with Prior stage, using Syncroscan RT software from Syncroscopy. Scaling for measurement was 182 nm/pixel and 5.5 pixels/micrometer at ϫ40 magnification.

Generation of GC2 Ϫ/Ϫ and GC1/GC2 Double Knock-out (GCdko) Mice-
The GC1 (Gucy2e, on mouse chromosome 11) and the GC2 (Gucy2f, on the X chromosome) genes are closely related in structure (Fig. 1, A and B) (34). The GC1 knock-out construct (18) replaced a portion of exon 5 encoding the transmembrane domain by a neo cassette, thereby deleting the entire intracellular region of GC1. We prepared a GC2 knock-out mouse by replacing a portion of exon 2 containing the translation start codon and the peptide leader sequence with a neo cassette (Fig. 1, B and C). Several generations of these mice were established on the C57BL/6J background and were cross-bred with GC1 Ϫ/Ϫ mice to generate GCdko mice. The GCdko mice were fertile and developed normally, thereby excluding a vital role for GC1 or GC2 during embryogenesis. The genotypes of single and double knock-out mice were confirmed by PCR amplification using GC1-and GC2-specific oligonucleotides (Fig. 1D). Immunoblots of GC1 Ϫ/Ϫ and GC2 Ϫ/Ϫ retinal lysates probed with monoclonal anti-GC1 and polyclonal anti-GC2 antibodies confirmed that GC1 and GC2 were not expressed (Fig. 1E). Further, immunoblots with anti-GC1 and anti-GC2 antibodies showed that the GC1 expression level is maintained in GC2 Ϫ/Ϫ retina and that the GC2 expression level is also maintained in GC1 Ϫ/Ϫ retina (Fig. 1F).
GCdko Rod and Cone Photoreceptors Are Physiologically "Silent"-To test for photoreceptor function in WT and single and double knock-out retinas, we performed full-field ERGs (Table 1 and Fig. 2). Consistent with a previous report (18), scotopic a-and b-wave amplitudes recorded from 4 -6week-old GC1 Ϫ/Ϫ mice were significantly reduced compared with WT responses (Fig. 2, A and B). However, scotopic ERG responses were absent in GCdko mice at all light intensities tested. These results suggest that the rod responses recorded from GC1 Ϫ/Ϫ retinas reflect GC2 activity. Scotopic ERG responses recorded from GC2 Ϫ/Ϫ mice closely resembled those of WT mice, implying that GC1 alone is able to support rod function under these conditions. The a-wave amplitude and sensitivity of the scotopic GC2 Ϫ/Ϫ mouse photoresponse were similar to WT at low intensities ( Fig. 2B; Table 1); however, at high intensities, GC2 Ϫ/Ϫ a-wave amplitudes were slightly reduced compared with WT (Fig. 2B). Photopic ERG responses recorded in the presence of background light were absent in GC1 Ϫ/Ϫ and GCdko mice, whereas GC2 Ϫ/Ϫ a-and b-waves were nearly identical to WT (Fig. 2, C and D). These results are consistent with GC1 regulating cGMP synthesis in cones and with both GC1 and GC2 regulating cGMP synthesis in rods. Given these observations, the possibility that other cyclases participate in rod or cone phototransduction can be excluded.
Recovery from Test Flashes as Determined by Paired Flash Analysis-A paired flash protocol was used to evaluate the ability of GC deletion mutant retinas to recover responsiveness after light stimulation (23,30,35). In this method, short test flashes followed by varying interstimulus intervals, and a probe flash were used to test for a-wave recovery (36,37). As expected, GCdko photoreceptors were unresponsive (Fig. 3A). Recoveries of a-waves measured from GC1 Ϫ/Ϫ mice and GC2 Ϫ/Ϫ mice were normalized to the amplitudes of corresponding WT a-wave responses. Rates of dark current recovery in GC1 Ϫ/Ϫ and GC2 Ϫ/Ϫ mice after the test flash were not significantly different from WT mice (GC1 Ϫ/Ϫ , 895.4 Ϯ 112.2 ms; GC2 Ϫ/Ϫ , 1018.1 Ϯ 63.1 ms; WT, 893.3 Ϯ 47.8 ms) ( Table 1).
Intense Bleach Delays Recovery of GC1 Ϫ/Ϫ and GC2 Ϫ/Ϫ a-wave Amplitudes-In a second approach, dark adaptation rates were determined by exposing animals to a high intensity bleaching light (500 cd⅐s⅐m Ϫ2 ) for 3 min and then recording a-wave responses elicited by probe flashes delivered intermittently over the course of a 60-min dark adaptation period (Fig.  3B). Under these conditions, 70% of rhodopsin was bleached (28). The ratios of the a-wave responses obtained during the recovery phase following the bleach for WT, GC1 Ϫ/Ϫ , GC2 Ϫ/Ϫ , and GCdko mice were plotted as a function of time in darkness. In WT mice, a-wave amplitudes recovered about 70% after 1 h of dark adaptation. The recovery of a-wave amplitudes was significantly delayed in GC1 Ϫ/Ϫ and GC2 Ϫ/Ϫ mice (p Ͻ 0.001) compared with WT. Following 1 h of dark adaptation, the amplitudes of the a-waves recorded from GC2 Ϫ/Ϫ and GC1 Ϫ/Ϫ mice had recovered to about 30 and 50%, respectively, of their dark-adapted values. These results suggest that GC1 and GC2 contribute distinctly to the recovery of photoreceptors from exposure to intense bleaching light.
Rod Photoreceptor Single Cell Recordings-The mean dark currents of GC1 Ϫ/Ϫ and GC2 Ϫ/Ϫ rod photoreceptors were not significantly different from the WT currents (12 Ϯ 1.2 pA for WT, 12.7 Ϯ 1.3 pA for GC1 Ϫ/Ϫ , and 13.8 Ϯ 1.3 pA for GC2 Ϫ/Ϫ ) (Fig. 4, A-C). In contrast, GCdko rods showed no detectable light-sensitive current (Fig. 4D). That cyclic nucleotidegated (CNG) channels are present at similar densities in the rod outer segments (ROS) of WT and GC single knock-out rods argues that [cGMP] i levels in darkness are similar for WT, GC1 Ϫ/Ϫ , and GC2 Ϫ/Ϫ rods, whereas effectively no cGMP is present in GCdko rods. It therefore follows that the basal activity of either GC1 or GC2 is sufficient to maintain dark current in rods.  (18). C, the knock-out construct of the mGC2 gene deletes a portion of exon 2 that includes ATG, the peptide leader sequence, and a portion of the mature N terminus of mGC2. D, PCR genotyping WT and knock-out GC1 gene (left) and genotyping the WT and knock-out GC2 gene (right). Diagnostic fragments for WT Gc1 (Gc1F4/G11R4) and knock-out Gc1 (NeoF4/Gc1R4) are 290 and 575 bp in size, respectively. Diagnostic fragments for WT GC2 (Gctowt1/Gctoda6) and knock-out GC2 (Pla2/Gc2mt-R2) are 1395 and 852 bp in size, respectively. E, immunoblots of retinal lysates probed with a polyclonal anti-GC1 antibody (left) and with a polyclonal anti-GC2 antibody (right). F, immunoblots of retinal lysates probed with polyclonal anti-GC1 and -GC2 antibodies. Anti-␤-actin antibody was used as an internal control for loading. Compensatory up-regulation of the other GC when one is missing may account for this observation; however, immunoblot results suggest normal GC2 levels in GC1 Ϫ/Ϫ mice (Fig. 1F) (18). Regarding the light response, the kinetics was broadly similar for all three genotypes (Fig. 4, A-C). Sensitivity of GC2 Ϫ/Ϫ rods was quite similar to that of WT rods (half-saturating flash intensity of 52.8 Ϯ 4.9 versus 48.6 Ϯ 6.3 photons m Ϫ2 ), but the sensitivity of GC1 Ϫ/Ϫ rods (half-saturating flash intensity of 13.6 Ϯ 2.5 photons m Ϫ2 ) was about 3.5-fold higher than WT (Fig. 4E) (see also Ref. 18).

Down-regulation of GCAP1, GCAP2, and Rod PDE6 in GCdko
Photoreceptors-Confocal immunolocalization was used to monitor the subcellular distribution of GCAPs and PDE6 following GC deletion (Fig. 5). As expected, GC1 was not detectable in GC1 Ϫ/Ϫ and GCdko sections (Fig. 5, B and D), and GC2 was undetectable in GC2 Ϫ/Ϫ and GCdko retinas (Fig.  5, G and H). GC1 deletion did not appear to affect GC2 expression, and GC2 deletion did not affect GC1 expression, in agreement with immunoblots (Fig. 1F). However, GC1 deletion strongly diminished GCAP1 levels in all outer segments (Fig. 5J) and GCAP2 levels in rods to a lesser extent (Fig. 5N). GC2 deletion had no obvious effect on GCAP1 or GCAP2 expression (Fig. 5, K and  O). In GCdko photoreceptors, low expression levels of GCAP1 and GCAP2 were confined to the cell bodies and myoid regions of photoreceptors (outer nuclear layer; ONL), where biosynthesis occurs (Fig. 5, L and P). Essentially no GCAPs were detectable in GCdko outer segments that, at this age, showed signs of early degeneration. The immunolabeling patterns of rhodopsin (Fig. S1, A-D) and transducin ␣-subunit (T␣; gene symbol Gnat1) (Fig. S1, E-H) were mostly normal. However, the labeling patterns of rhodopsin kinase (Fig. 6T) and rod transducin ␣-subunit (Fig. S1H) were consistently "patchy" over GCdko ROS.

FIGURE 2. Full field ERGs of WT and GC deletion mutant mice under scotopic and photopic conditions.
A, serial responses to increasing intensities of flash stimuli were obtained from dark-adapted WT, GC1 Ϫ/Ϫ , GC2 Ϫ/Ϫ , and GCdko mice. Although the GC1 Ϫ/Ϫ response was severely attenuated, only double knock-out mice failed to produce any ERG responses. B, the a-and b-wave amplitudes, respectively, plotted as a function of light intensity under dark-adapted conditions. C, three representative serial responses to increasing intensities of flash stimuli were obtained from light-adapted WT, GC1 Ϫ/Ϫ , GC2 Ϫ/Ϫ , and GCdko mice. D, a-and b-wave amplitudes plotted as a function of light intensity under light-adapted conditions. 4 -6 littermates of each genotype were used for analysis.
suggested that cone PDE6␣Ј was also down-regulated in the GCdko mutant. GC1 Ϫ/Ϫ and GCdko Cone Outer Segments Are Unstable-Previous studies showed that GC1 Ϫ/Ϫ cone photoreceptors develop and elaborate outer segments but gradually degenerate (18,38). The subcellular distributions of S-and M/L-pigments that reside exclusively in cone outer segments (COS) and the freely diffusible cone arrestin were investigated in the three GC deletion mutants by immunocytochemistry. S-opsin (Fig. 6, A-D) and M/L-opsin (Fig. S2), which normally reside in COS, were present at reduced levels in the GC1 Ϫ/Ϫ and GCdko cones and mislocalized. Extracellular particles, intensely immunopositive for cone pigments, were observed adjacent to the retinal pigment epithelium (arrows in Fig. 6, B and D, and Fig. S2B; see also Fig. 8D). Cone somata were still abundant at 2 months of age in GC1 Ϫ/Ϫ retinas but with highly disorganized COS (Figs. S2, B and F, and 8, C and E), and GCdko retinas had a slightly reduced number of cones as judged by labeling with anti-cone arrestin (Fig. S2H). Double labeling with anti-cone arrestin and anti-S-pigment antibodies showed that S-pigment was present in GC1 Ϫ/Ϫ outer segments predominantly in extracellular particles (Fig. S2J, arrows). GC1 Ϫ/Ϫ and GCdko COS are apparently unstable and disintegrating at the age of 2 months.
Cone Transducin, Cone PDE6␣Ј, and GRK1 Are Undetectable in GC1 Ϫ/Ϫ and GCdko Cone Outer Segments-We next tested for the presence of membrane-associated cone phototransduction proteins in GC mutant COS. Specifically, we examined levels of myristoylated cone T␣, farnesylated cone T␥, geranylgeranylated cone PDE6␣Ј, and farnesylated GRK1 in GC1 Ϫ/Ϫ , GC2 Ϫ/Ϫ , and GCdko retinas (Fig. 6). All four proteins were undetectable in GC1 Ϫ/Ϫ (Fig. 6, F, J, N, and R) and GCdko COS (Fig. 6, H, L, P, and T). Note that GRK1 immunolabeling distinguishes between rods and cones and was comparable with that of WT in GC1 Ϫ/Ϫ and GC2 Ϫ/Ϫ ROS (Fig. 6, Q-S) but was undetectable in GC1 Ϫ/Ϫ and GCdko COS (Fig. 6, R and T). These results show that essential phototransduction proteins The dark-adapted mice were conditioned first with the test flash (0.4 log⅐cd⅐s⅐m Ϫ2 ) followed by a probe flash (1.6 log⅐cd⅐s⅐m Ϫ2 ) with interstimulus intervals ranging from 200 to 2,000 ms. Each trace represents the average recording from eight eyes. The interstimulus interval required to reach 50% a-wave amplitude recovery was not significantly different among genetic backgrounds ( Table 1). As expected, GCdko photoreceptors were unresponsive. B, measurements of a-wave recovery following constant light stimulation. Dark-adapted mice were bleached with intense constant illumination (500 cd⅐s⅐m Ϫ2 ) for 3 min, and recovery of a-wave amplitude was monitored by recording single-flash ERGs (Ϫ0.2 log⅐cd⅐s⅐m Ϫ2 ) over the course of a 60-min dark adaptation period; the recovery ratio is plotted. The ratio in GC1 Ϫ/Ϫ , GC2 Ϫ/Ϫ , and GCdko mice was significantly attenuated compared with WT (GC1 Ϫ/Ϫ and GC2 Ϫ/Ϫ compared with WT, p Ͻ 0.001; GCdko compared with WT, p Ͻ 0.0001), and the ratio in GC2Ϫ/Ϫ was significantly attenuated relative to GC1 Ϫ/Ϫ mice (GC1 Ϫ/Ϫ to GC2 Ϫ/Ϫ , p Ͻ 0.05). cone T␣, cone T␥, PDE6␣Ј, and GRK1 are either absent or below the threshold of detection in GC1 Ϫ/Ϫ COS.
Down-regulation of Membrane-associated Proteins Is Posttranslational-By immunocytochemistry, we observed significant down-regulation of GCAPs, cone T␣, cone T␥, and cone PDE6␣Ј in GC1 Ϫ/Ϫ and GCdko photoreceptor outer segments, whereas GC2 deletion had no effect on protein levels. Immunoblots (Fig. 7A) confirmed that protein levels of GCAPs, cone T␣, and cone arrestin were in part drastically reduced in GC1 Ϫ/Ϫ and GCdko retinas. Absence of rod PDE6 in GCdko rods was also confirmed by immunoblotting; rod PDE6 subunits (␣, ␤, and ␥) were undetectable in GCdko retina extracts. In contrast, deletion of either GC singly had no effect on rod PDE6 expression. Rhodopsin, rod T␣, and rod CNG channel subunit ␣ subunit (CNGA1) were present at nearly normal levels in all four genotypes.
Down-regulation of GCAPs was shown to be posttranslational in GC1 Ϫ/Ϫ retina (19). We extend that observation to show that other peripherally associated membrane proteins are also down-regulated. To verify that transcription of genes whose products are down-regulated in GCdko retina is normal, we determined the concentrations of mutant mRNAs relative to WT mRNAs by semiquantitative real time RT-PCR (Fig. 7B). Real time RT-PCR analyses of GCAP1, GCAP2, rod PDE6␣, and cone T␣ transcripts confirmed that mRNA levels were essentially identical in both WT and double knockout retinas. A standard curve indicated similar amplification kinetics for each transcript (Fig. S3). Microarray analysis suggested that the transcript levels of the downregulated genes (and all other phototransduction genes) in GCdko retina were essentially unchanged (results not shown). Thus, genes encoding GCAPs, rod/cone PDE6, and cone transducin are transcribed normally in the GCdko retina, and down-regulation occurs posttranslationally.
GCdko Retinas Degenerate as in Human LCA-At 2 months of age, outer segment lengths are ϳ50 -70% of normal, suggesting that degeneration progresses relatively slowly in GCdko mice (e.g. Fig. 5, D,  H, and L). We observed that cone degeneration was consistently most severe inferior to the optic nerve (results not shown); the superior midperiphery was consistently the least degenerated as observed by Coleman et al. (19). At 6 months of age, the GCdko outer nuclear layer contains only 4 -6 rows of nuclei (Fig. S4B, right), and ROS lengths are severely reduced in superior/inferior and nasal/temporal quadrants (Fig. S4A). Cone cell remnants, particularly synaptic pedicles and somata, can be identified by labeling with anticone arrestin (Fig. S4B, bottom), but outer segments are absent. This finding is consistent with a post-mortem study of an 11-year-old LCA patient with a null mutation in the GUCY2D gene. The macula and peripheral retina of this patient revealed a substantial number of cones and rods, but both photoreceptor types lacked outer segments (39).
Ultrastructural Examination Reveals Membrane Partitioning in GC1 Ϫ/Ϫ and GCdko Photoreceptors-To provide a structural framework for the interpretation of immunolocalization, companion eyes were plastic-embedded and sectioned for electron microscopy. GC1 Ϫ/Ϫ retinas were selected for the examination of cones, GCdko retinas were selected for analysis of rods, and each of these was compared with the corresponding photoreceptor of an age-matched WT retina. All observations are derived from mouse retina inferior to the optic nerve, a region where electron-dense granules were frequently encountered in distal inner segments of both WT and knock-out cones (see Fig. 8B). GC1 Ϫ/Ϫ cones exhibited problems of varying degree; apical inner segments appeared retracted from connecting cilia and surrounded by extracellular "blebs"; at the distal connecting cilia, membrane components appeared to partition into layered stacks or vesicles/tubules, and in severely degenerated cases, these membranes were detached (Figs. 8,  C and E). Adjacent to the retinal pigmented epithelium, particles consisting of an inner collection of vesicles enclosed by layers of dense membrane were often observed (e.g. see Fig. 8D) and correspond in position to cone opsin-immunoreactive particles of Fig. 6B. GCdko rods, too, were surrounded by blebs in the interphotoreceptor matrix. Distal to the connecting cilia, the ROS were shorter and narrower than their WT counterparts and appeared "banded" with alternating regions of dense membrane layers and lumen-containing tubules (Fig.  8, G and H) suggesting unconventional ROS membrane assembly.

DISCUSSION
Deletion of integral membrane proteins GC1 and GC2, singly and in combination, has permitted the following observations in mouse photoreceptors. First, GC2 can compensate in part for the absence of GC1 in rods. Second, simultaneous ablation of GC1 and GC2 abolishes rod and cone phototransduction, thereby excluding the possibility that a third GC (soluble or particulate) participates in the regulation of phototransduction. Third, several peripherally membrane-associated phototransduction components (GCAPs, rod and cone PDE6, and cone transducin) are either absent or severely reduced in GC1 Ϫ/Ϫ and GCdko photoreceptor outer segments. Fourth, GC1 Ϫ/Ϫ and GCdko cone outer segment membranes are disorganized at 1-2 months. Fifth, GCdko rod and cone photoreceptors degenerate with a phenotype simulating human LCA1 (for a phenotype summary of GC knock-outs, see Table S1).
Apart from a minor ERG a-wave amplitude reduction, GC2 Ϫ/Ϫ rod photoreceptors recovered more slowly both in double flash experiments (Fig. 3A) and after intense 3-min photobleaching (Fig. 3B). The mechanism for the delay is unclear, but it may in part be caused by a reduction in GRK1 (Fig. 7A) (40,41). In single cell recordings, GC2 Ϫ/Ϫ rods and WT rods displayed nearly identical half-maximal sensitivity. However, return to the dark state in GC2 Ϫ/Ϫ rods expressing was slightly FIGURE 6. Immunolocalization of S-opsin, cone T␣, cone T␥, cone PDE6␣, and GRK1 in retinas of WT and GC deletion mutants, aged 1-2 months. A-D, localization of S-opsin; E-H, cone T␣; I-L, cone T␥, M-P, cone PDE6␣Ј; Q-T, GRK1 in WT, GC1 Ϫ/Ϫ , GC2 Ϫ/Ϫ , and GCdko retinas, respectively. S-opsin, cone T␣, cone T␥, cone PDE6␣Ј, and GRK1 were localized normally to WT and GC2 Ϫ/Ϫ COS. S-opsin mislocalizes to COS fragments in GC1 Ϫ/Ϫ and GCdko retinas. Cone T␣, cone T␥, and cone PDE6␣Ј appear down-regulated in GC1 Ϫ/Ϫ and GCdko retinas. Anti-GRK1 labels cone outer segments prominently in the proximal OS layer, whereas the remainder is attributable to ROS. GRK1 is present in ROS of all four genotypes and in WT and GC2 Ϫ/Ϫ COS but absent in GC1 Ϫ/Ϫ and GCdko COS. Magnification bars, 10 m.  A, open arrows denote connecting cilia of cone photoreceptors for reference. The WT cone outer segment measures 1.3 m at its widest point and tapers distally; B, cone inner segments were sometimes observed to contain electron-dense granules (stars). C, disorganized GC1 Ϫ/Ϫ cone outer segment reveals apparent partitioning of membrane components; lamellar structures emanate from one side of the distal connecting cilium and vesicles from the tip. D, adjacent to the retinal pigmented epithelium (RPE), extracellular particle (2-3 m in diameter) containing vesicles and tubules surrounded by membrane lamellae appears as an end product of membrane component sorting. Such a particle probably corresponds to the opsin-immunoreactive particles shown in Fig. 6B (see also Figs. S2B and S2J). E, another example of membrane partitioning is shown in which membrane layers are found to one side of the distal connecting cilium and vesicles/ tubules extend from its tip. Cones were examined in a location inferior to the optic nerve for each genotype. F-H, electron micrographs of rod outer segments from WT and GCdko mouse retinas. The open arrow indicates the connecting cilium of the WT rod (F). Predegenerate ROS of GCdko retinas (G and H) have a striated appearance of alternating lamellae and tubules that may underlie the patchy patterns of rhodopsin kinase immunoreactivity observed in Fig. 6T and rod T␣ immunoreactivity in Fig. S1H. MARCH 23, 2007 • VOLUME 282 • NUMBER 12 time RT-PCR (Fig. 7B) and microarray analysis (not shown). Posttranslational down-regulation of these proteins may result from several mechanisms, including proteolytic degradation, protein instability, and impeded vesicular transport. GCAPs have been shown to be susceptible to proteolysis in low free Ca 2ϩ in vitro (43) conditions presumed to be predominant in the GCdko and GC1 Ϫ/Ϫ photoreceptors in which cation channels are closed. Therefore, proteolytic degradation of GCAPs in low [Ca 2ϩ ] could provide a mechanism for GCAP down-regulation. However, transducin (44) and PDE6 subunits (45) can be expressed stably in tissue culture in the absence of GC, cGMP, or external Ca 2ϩ . It seems unlikely that low occupation of noncatalytic cGMP binding sites (GAF domains) (46) present on catalytic subunits influences PDE6 stability.

Deletion of GC1 and GC2 in Retinal Photoreceptors
Common to all down-regulated proteins is their membrane association, either directly via prenyl or acyl anchors or indirectly by interaction with integral membrane proteins (GCAPs with GCs). Peripheral membrane-associated proteins are firmly bound to the membrane surface, are nondiffusible, and therefore must be transported to the site of ROS membrane assembly. Mechanistic details of transport of membrane-associated proteins are unknown. A vesicular transport pathway for rhodopsin from the endoplasmic reticulum to the outer segments has been firmly established (47)(48)(49)(50). Transmembrane GCs probably follow a similar pathway. Outer segments of mouse photoreceptors undergo renewal about every 10 days (51), a high turnover rate requiring active biosynthesis and efficient transport. Thus, transport and targeting of membrane proteins to the outer segments is a pathway that is essential for function and survival of photoreceptors. We surmise that peripheral membrane proteins may attach to vesicular transport carriers for transport to the outer segments.
Knock-out of GCs does not affect the rhodopsin transport pathway, because rhodopsin is found at nearly normal levels in GCdko ROS (Fig. 7A, S1D). Therefore, GC transport and rhodopsin transport in rods must occur independently. Deletion of one GC does not affect stability and/or transport of PDE6 subunits in rods, but deletion of both GCs leads to PDE6 removal as evidenced by the absence of PDE6 signal on immunoblots (Fig. 7A). The presence of at least one GC is essential for PDE6 survival.
Knock-out of GC1, the only GC present in cones, strongly affects levels of S-cone and M/L-cone pigments (Figs. 6B and S2B). Both pigments localized to membranes that appear in part detached from the cone inner segments in GC1 and GCdko mutants. A second distinction between rods and cones is that acylated cone T␣ and farnesylated cone T␥ were undetected in GC1 Ϫ/Ϫ COS (Fig. 6, F and J), whereas acylated rod T␣ (Fig.  S1H) and farnesylated rod T␥ (results not shown) distributed normally in GC1 Ϫ/Ϫ and GCdko rods. Third, GRK1 was undetectable in GC1 Ϫ/Ϫ and GCdko COS but present in ROS. Thus, GC deletions affect rod and cone photoreceptor outer segments differently, suggesting that separate pathways have evolved in rods and cones for transport of membrane proteins. GC1 deletion only affects cones, whereas GC2 deletion has no effect on distribution of membrane-associated proteins. Deletion of both GCs has the additional effect of preventing PDE6 transport to the ROS.
It is noteworthy that in the absence of GCs, rod and cone outer segment membranes are assembled unconventionally, become unstable, and eventually disintegrate. Electron microscopy of GC knock-outs (Fig. 8) reveals an apparent "partitioning" of membrane components such that GCdko rod outer segments are composed of alternating bands of dense membrane stacks and lumen-containing tubules within a plasma membrane, whereas GC1 Ϫ/Ϫ cone outer segments are composed of irregular collections of vesicle/tubules segregated from dense membrane layers. Conceivably, as transmembrane proteins, the presence of GCs may provide a structural scaffold that supports stabilization of disk morphogenesis by integrating membrane components. Further, the GC deletion phenotype is predicted to have low levels of cytoplasmic cGMP, resulting in permanently closed CNG cation channels and low [Ca 2ϩ ] i , a situation mimicking constant light without phototransduction signaling. Mutations mimicking constant light ("equivalent light hypothesis") have been proposed to trigger retinal degenerations in several, diverse animal models (52).
The GCdko retina degeneration phenotype is characterized by low levels of cGMP and is accompanied by a rod PDE6 knockdown. This suggests that the recessive retinal degeneration in the rd1 mouse (53,54), which has high levels of cGMP due to an incapacitated PDE6, cannot be rescued by deletion of GCs.
In summary, our genetic, biochemical, and physiological approach to elucidate the function of GCs in mouse rod and cone photoreceptors revealed that GC1 and GC2 are the only cyclases involved in phototransduction. We have found no evidence for a role of soluble GCs or a third particulate GC as is known to exist in fish (55). GC1 appears to be a dominant enzyme, particularly in cones, but GC2 can effectively produce cGMP in the absence of GC1 in rods. Deletion of both GCs results in an animal model without rod and cone vision and a phenotype of a recessive cone/rod dystrophy resembling human Leber congenital amaurosis.